Resources
The Resources provides detailed specifications for high-speed and ultra-high-speed training GPU models, along with key advantages such as powerful computing, high network performance, elastic expansion, and seamless integration with deep learning environments.
Product Specifications
| Resource Type | GPU Model | GPU Memory | Number of GPU Blocks | Training Type |
|---|---|---|---|---|
| High-speed training | NVIDIA 4090 | 24G | Single card/whole machine | High-speed training |
| High-speed training | NVIDIA 2080 | 8G | Single card/whole machine | High-speed training |
| Ultra-high speed training | NVIDIA A100 | 40/80G | Single card/whole machine | Ultra-high speed training |
| Ultra-high speed training | NVIDIA A800 | 80G | Single card/whole machine | Ultra-high speed training |
| Ultra-high speed training | NVIDIA H800 | 80G | Single card/whole machine | Ultra-high speed training |
| High-speed training | NVIDIA A10 | 24G | Single card/whole machine | High-speed training |
| High-speed training | NVIDIA L40 | 4/8G | Single card/whole machine | High-speed training |
Resource advantages
Ultimate Computing Power
NVIDIA graphics cards, combined with high-performance CPUs, provide users with powerful computing power.
High Network Performance
Based on the Mellanox HDR InfiniBand solution, it supports InfiniBand unlimited bandwidth technology and has extremely high throughput and extremely low latency.
Perfect Supporting Services
The server uses NVMe storage devices locally, supports mounting all-flash parallel file storage, and has exclusive Internet export bandwidth.
Elastic Expansion
As user business grows, the AI computing resources on the cloud can be flexibly increased without interrupting business.
Deep Learning Environment Integration
Integrate the k8s scheduling system. The public image integrates basic functions such as CUDA and Python, as well as deep learning frameworks such as TensorFlow and PyTorch.